docker_cuda10.1

Download the image:

docker pull tensorflow/tensorflow:nightly-custom-op-gpu-ubuntu16-cuda10.1

Run/Create the image you downloaded:

docker run tensorflow/tensorflow:nightly-custom-op-gpu-ubuntu16-cuda10.1

Docker 快速教學

https://blog.gtwang.org/virtualization/ubuntu-linux-install-docker-tutorial/

Remote the docker with simple GUI:

Download: docker pull jarkt/docker-remote-api
Run: docker run -p 2375:2375 -v /var/run/docker.sock:/var/run/docker.sock --name docker-remote-api jarkt/docker-remote-api
(it requires the terminal keep opening)

Open your browser and verify you can connect to http://<Your IP that runing docker>:2375/_ping

Start/Stop:
Start : docker start docker-remote-api
Stop : docker stop docker-remote-api

Build the environment

In this example, we are using ubuntu16.04.
Go download and install some essential components. -> CUDA10.1

Start session

-i, --interactive Attach container's STDIN
docker start -i <dockerID>
for example:
docker start -i 45c08877c88a